Glossary
Hallucination — Glossary
When a generative model produces fluent content that is fabricated, unsupported, or inconsistent with provided sources.
Full definition
In AI discussions, “hallucination” describes confident but incorrect model outputs—invented facts, bogus citations, or details not present in evidence. The term is metaphorical; models are not perceiving imaginary objects the way a person might.
Hallucinations arise because generators optimize for plausible continuations, not guaranteed truth. Mitigations include retrieval grounding, tool use, citation checks, constrained decoding, lower-stakes use cases, and human review. Measuring hallucination rates on domain-specific tests is more informative than relying on anecdotes.